| Breeding industry is a pillar industry in southern Xinjiang.Such external environmental factors inevitably affect the internal environment of chicken houses.In recent years,the breeding of laying hens in southern Xinjiang has gradually developed towards intensification and scale.Intensive farming has brought efficient production,but high density,large-scale farming also highlights the environmental problems of intensive hen houses.The traditional environmental control methods of laying hens mostly rely on a certain environmental parameter to control,and the coupling of various environmental factors cannot achieve the best control effect,thus affecting the growth and production performance of laying hens.Aiming at the problem that the traditional control method can not effectively deal with a variety of environmental factors,the environmental monitoring system of a laying hen house in southern Xinjiang was designed and implemented.The reinforcement learning method was applied to the environmental control model of the laying hen house.The temperature and humidity index and the prior knowledge were introduced to optimize the reinforcement learning method.The optimal control model of the laying hen house in southern Xinjiang was obtained.The experimental results show that the optimal control model can smooth the environmental parameters of the laying hen house and improve the egg production rate of the egg.This study includes the following aspects:（1）The environmental monitoring system of laying hens house is designed and implemented based on Internet of Things technology.In view of various factors affecting the growth environment of laying hens,the monitoring function of various environmental parameters of laying hens was realized,and the control function of environmental control equipment such as fans and water curtains in laying hens was realized.The environmental monitoring platform of laying hens was built,and finally the environmental monitoring system of laying hens was realized.During the experiment,2749 pieces of actual data were collected,2772pieces of data should be collected,23 pieces of data lost,and the data acquisition rate was 99.18%,which met the requirements of system design.The correlation among dry bulb temperature,relative humidity and carbon dioxide concentration of chicken house environmental factors was analyzed.The results showed that the three indexes were correlated with each other and the correlation was obvious.The design and implementation of the laying hen house environment monitoring system provides a research basis for the control of the laying hen house environment.（2）Laying hen house environment control model based on reinforcement learning method.Firstly,four environmental factors including temperature,relative humidity,ammonia concentration and carbon dioxide concentration in the hen house were selected as state parameters by reinforcement learning method,and each state parameter was divided according to the region.Secondly,the switches of fans and water curtains in the hen house are combined as action parameters.At the same time,the reward value function of the hen house environment is established according to the relevant standards and the actual situation of the hen house,and the environmental control model of the hen house based on reinforcement learning method is completed.Finally,the environmental control model of southern Xinjiang laying hens was compared with the threshold switching control method,and the effectiveness of the control model was verified.Through the actual verification,based on the reinforcement learning environment control model,the temperature of the hen house was maintained at 19.1℃-20.6℃,the relative humidity was maintained at 53%-69.2%,and the average carbon dioxide concentration（5751mg/m~3）was also greatly improved compared with the threshold switch control（6963mg/m~3）.During the experiment,the laying rate（74.2%）of the laying hens under the reinforcement learning control model was also higher than that under the threshold switch control（72.6%）.The experimental results showed that the environmental control model of laying hens house based on reinforcement learning could effectively realize the stable control of various environmental factors in the laying hens house in southern Xinjiang,so as to ensure the growth and production performance of laying hens.（3）Optimization of environmental control model of laying hens in southern Xinjiang.The model takes the temperature and humidity index as the state parameter of reinforcement learning and the prior knowledge as the action decision to optimize the control model.A)Based on the collected temperature and relative humidity,the temperature-humidity index model is constructed to replace the two parameters of temperature and relative humidity in the reinforcement learning model,thereby reducing the dimension of state parameters.B)Using prior knowledge as action decision-making,abandoning learning in an environment where chickens are obviously not adapted,and choosing pre-set actions to avoid unnecessary environmental impacts on chickens in the learning process and reduce the number of learning.The experiment showed that the number of times to achieve the convergence of the model was significantly reduced from 1500 times to about 620 times,and the laying hens were in the comfort zone of the temperature and humidity index during the experiment,and did not reach the heat stress level.In this study,based on the data collected by the egg-laying hen house environmental monitoring system,the egg-laying hen house environmental control strategy was studied by reinforcement learning method,and the optimal control model of the egg-laying hen house environment in southern Xinjiang was established.The research results achieved the smooth control of multiple environmental factors in the laying hens‘house,so as to improve the growth and production environment of laying hens and increase the egg production rate,which had certain feasibility in the control of laying hens‘house. |
PDF Full Text Request